How to make a second ad from the same footage that reads as a different creative

Last updated 2026-09-10 · 8 min read

What is the short version?

Testing two versions only tells you something if a viewer would recognise them as two separate ads. Reordering the same shots produces a variant that performs almost exactly like the original, because to anybody watching it simply is the original in a different order.

1. Go to Vizard Agent with the first version.
2. Say the second must read as a different creative.
3. Ask for different takes and a different cutting style.

What do you need before you start?

The source and the first version. Vizard Agent builds the second from the same material, so it needs to know what the first one used — otherwise the two overlap by accident in exactly the shots that were strongest.

• The full source. Not just the first cut.
• The first version. So it can differ from it deliberately.
• The length. Often different from the first.
• What must appear in both. The product, usually.
• Whether the audio is live. It changes the assembly.

What do you type into Vizard Agent?

Tell Vizard Agent what "different" actually means to you here. Left open, it can mean a new order, a new length or a new look, and the version that genuinely tests something against the first usually needs all three of those at once.

Prompt

Variants worth knowing:

• "A separate creative, not a recut." The actual brief.
• "Change the cutting style as well." The third variable.
• "Keep each scene's original sound." Stops a borrowed track.

What does Vizard Agent actually do?

Here is the sequence Vizard Agent worked through on a real product ad that needed a second creative for testing. Notice how much of it is about the audio, which is the part that quietly breaks the moment shots are put into a different order.

1. Checks which takes are available and how long each runs.
2. Reads the original scene map for the exact audio times.
3. Locates the audio interval belonging to each take.
4. Chooses a different set of takes from the first version.
5. Puts them in a different order.
6. Builds at the new length.
7. Reorganises the original audio so each scene keeps its own sound.
8. Changes the cutting grammar — hard cuts rather than dissolves.
9. Checks the waveform for gaps or silence after the rebuild.
10. Reviews the assembly for variety of scenes and framing.

Step three is the one that makes live sound survive a reorder. Real audio belongs to a moment, not to a timeline, so moving a shot without bringing its sound produces a cut where the picture and the noise are from different seconds.

Step eight is what separates a second creative from a second edit. Order and shot choice change what is shown; cutting grammar changes how it feels, and a version built from hard cuts reads as a different piece of work even where it shares a shot with the first.

Step nine catches the standard failure of a rebuild. Reassembled audio is exactly where a silent gap or a format mismatch appears, and it is inaudible on a phone speaker until somebody watches it properly.

What does the result look like?

Two ads from one shoot that a viewer would never identify as the same edit. They share a product and roughly a length bracket and almost nothing else, which is exactly what makes a comparison between the two of them mean anything at all.

Vizard Agent also tells you which takes each version used, so a third variant can be built without repeating either of them.

When does this not work well?

Two genuinely different creatives need enough material to make two, and Vizard Agent will say when there is not. A short shoot with four usable takes cannot produce two versions that avoid overlapping, and forcing it gives you a weak second cut rather than a different one.

• Thin coverage. Not enough takes to go round.
• One hero shot. Both versions will want it.
• A fixed script. The words tie the order down.
• Very short formats. Six seconds has no room to differ.
• Strict brand templates. The grammar may be prescribed.

How do you fix a result that came back wrong?

Say which of the dimensions is not different enough yet. Because Vizard Agent varied the takes, the order, the length and the cutting grammar as separate decisions, naming the one that still feels the same as version one is a specific instruction rather than a request for another attempt.

• "It still feels the same." Change the cutting grammar.
• "You reused the best shot." Reserved for one version.
• "The audio does not match the picture." Re-associate per take.
• "Too different — keep the opening." One shared beat.

How does Vizard Agent compare to doing it yourself?

By hand a second version is made by opening the first and moving things, because the timeline is already built. That is why most test variants are recuts — the tooling makes reordering easy and starting again expensive.

Common questions

Why not just reorder? Because a viewer recognises the same shots, so the test compares nothing.

How different is different enough? Different takes and a different cutting style. Vizard Agent will show you both.

Can it avoid reusing the best shot? Yes. Tell Vizard Agent to reserve it for one version.

What about the audio? Vizard Agent keeps each take's own sound with it rather than one track.

Can the lengths differ? Yes, and usually they should.

How many versions can it make? As many as the coverage supports; Vizard Agent will say when it runs out.

Does the product have to appear in both? Almost always. Tell Vizard Agent what is mandatory in every version.

Will it change the music? Vizard Agent can, and a different bed helps a great deal here.

What if there is a voiceover? A fixed script limits the order. Vizard Agent will tell you how much.

Can it list what each version used? Yes. Vizard Agent reports it, which makes a third variant straightforward.

Does it check the audio after rebuilding? Yes, on the waveform, where reassembly faults appear.

Should both versions be the same length? Not necessarily. Different lengths are part of the test.

Can it copy a competitor's grammar? It can work from a reference you supply for the cutting style.

Why do my variants perform the same? Because they are the same ad in a different order, and viewers see that.